1st draft - university of liverpoollivrepository.liverpool.ac.uk/1125/1/1125.doc · web...
TRANSCRIPT
D Buck
The development and validation of the Epilepsy and Learning Disability Quality
of Life (ELDQOL) scale
Deborah Buck1, Monica Smith1, Richard Appleton2, Gus A Baker3, Ann Jacoby4
1 Institute of Society and Health, Newcastle University, UK2 Alder Hey Children’s Hospital, Liverpool, UK3 The Walton Centre for Neurology and Neurosurgery, Liverpool, UK4 Dept of Public Health, University of Liverpool, UK
Corresponding author:
Professor Ann Jacoby
Dept of Public Health
Whelan Building
University of Liverpool
Brownlow Hill
Liverpool
UK
Phone: +44 (0) 151 794 5602
Fax: +44 (0) 151 794 5588
Email: [email protected]
Word count: 2,646 (excluding tables and references)
Keywords: epilepsy, children, learning disabilities, quality of life, carer, assessment
1
D Buck
Abstract
Few suitable instruments exist for use with people, especially children, with both
epilepsy and learning disabilities. One such measure is the Epilepsy and Learning
Disabilities Quality of Life scale (ELDQOL), which has recently undergone revision
following feedback from relevant users. This paper reports on the final psychometric
testing phase of ELDQOL. ELDQOL consists of 70 items covering seizure severity,
seizure-related injuries, AED side-effects, behaviour, mood, physical, cognitive and
social functioning, parental concern, communication, overall QOL and overall health.
Revalidation involved a qualitative phase to ascertain users’ opinions on the wording,
coverage and layout of the questionnaire; and a quantitative phase to examine internal
consistency and test-retest reliability, and validity. The final version of ELDQOL has
very good evidence of reliability and validity, making it a promising instrument for
assessing QOL in children/young adults with epilepsy and learning disability.
Word Count: 138
2
D Buck
Introduction
There is general agreement within the health outcomes field of the importance of
assessing therapeutic outcomes including quality of life (QOL).1 Medical
interventions may be beneficial to patients on impairment or disability measures, but
without equally refined QOL measurements a clear and comprehensive evaluation of
their efficacy is not possible.2,3 The assessment of QOL in adults with epilepsy is now
widespread with the existence of a number of psychometrically sound measures.4
However, reviews of outcome measures have shown that few suitable instruments
currently exist for use with people who have both epilepsy and learning disabilities.1,5
One recently developed measure, the Epilepsy Outcome Scale (EOS),6 was developed
to be completed by carers of people with epilepsy and learning disabilities (LD), but
while valid and reliable, it is intended for adults rather than children with these
conditions. A recent review of QOL measures for use specifically with children and
adolescents with epilepsy1 identified five epilepsy-specific instruments, of which three
were considered potentially suitable for children who had LD: the Health-related
Quality of Life in Children with Epilepsy measure; 7 the Impact of Childhood
Neurological Disability Scale (ICND); 8 and the Epilepsy and Learning Disabilities
Quality of Life scale (ELDQOL).9-12
The Health-related Quality of Life in Children with Epilepsy measure was not
purposely developed for those with both epilepsy and LD. However, it does
incorporate a self-report (completed by the child) as well as a proxy-report which can
be completed by parents alongside or instead of the child report, and seems to be
reasonably valid and reliable.1 The ICND assesses the impact of epilepsy and
concomitant behavioural, cognitive, and physical/neurological disability in children. It
has good reliability and is reported to be satisfactory in terms of validity.1
ELDQOL is specifically aimed at informal/formal carers of children with both severe
epilepsy and LD. Initial evidence of its psychometric properties was reported by the
scale’s developers13, and subsequently ELDQOL was reviewed by Espie et al5 who
also reported evidence of internal and test-retest reliability, and some evidence of
sensitivity. ELDQOL has been used in two studies, results of which have been
3
D Buck
reported.11,14
However, as a result of feedback from clinicians involved, it was decided to undertake
revision and revalidation of ELDQOL. This paper reports on the psychometric testing
of the final, revised version of ELDQOL. A detailed report of the earlier stages in its
development is available from one of the authors (AJ). A summary of its history will
be provided here.
Initial development of ELDQOL
Potential items for inclusion in the pilot structured measure were selected based on in-
depth interviews with parents of children with severe epilepsy. The measure was then
piloted to assess its psychometric properties (content and construct validity, internal
consistency and test-retest reliability) and acceptability to parents. Parents of 50
patients across 5 UK paediatric neurology out-patient departments were recruited to
the pilot study.
The pilot version of ELDQOL consisted of 66 items across 6 domains: seizure
severity, seizure-related injuries, AED side-effects, behaviour, mood and overall QOL.
It was found to be satisfactory in terms of content and construct validity and internal
consistency, and showed high test-retest reliability. Furthermore, parents commented
that the measure was comprehensive, relevant and easy to complete. The pilot version
of ELDQOL was subsequently used in a clinical trial of Lamictal11 which provided
evidence of its responsiveness. However, a number of criticisms were made about the
measure by clinicians involved in the trials, particularly in terms of the clarity of
meaning of several items and of the appropriateness or sensitivity of response sets.
Consequently, a revalidation study was undertaken to pre-test (using cognitive
interviews15), revise and then reassess the psychometric properties of ELDQOL. The
purpose of this paper is to describe this final phase in the development of the measure
and report its psychometric properties.
4
D Buck
Methods
Qualitative phase
Qualitative (cognitive) interviews were undertaken with 16 parents and 17 health and
other professionals in order to ascertain their opinions on the wording, coverage and
layout of the questionnaire.
The parent participant group comprised 5 couples (the young person’s mother and
father) and 6 mothers. The 11 children of these parents were 8 boys and 3 girls, age
range 18 months to 21 years. Two young people attended residential school, 7
attended special schools or day centres, and one a mainstream school. The parents
were identified by a consultant or senior nurse at 2 specialist hospitals in the north of
England. They were selected because of their child’s condition and because they were
judged to be interested in all aspects of their care.
The professional participant group consisted of 10 consultant paediatric neurologists
or neurologists with a special interest in epilepsy and learning disabilities, one
associated health services researcher, two paediatric epilepsy nurse specialists, two
nurses specialising in epilepsy and learning disabilities, two teachers/play leaders, and
one psychologist.
All participants were sent a copy of the existing ELDQOL questionnaire and asked to
complete it immediately prior to the in-depth interview. Interviews were tape
recorded and lasted approximately one hour. They involved scrutiny of each question
in turn, where participants’ views were sought on interpretation, coverage, content,
wording and scale construction. The tapes were transcribed for analysis.
Psychometric phase
Following revision based on the findings of the cognitive interviews, the final version
of ELDQOL was administered in a postal survey to 47 parents/guardians and 21
formal carers of children with epilepsy and learning disabilities. The parents were
identified by consultant paediatric neurologists and epilepsy specialist nurses at 4 UK
5
D Buck
hospitals, who explained the study and asked whether they would be willing to
complete the questionnaire. Those who agreed were sent a study information sheet
together with a copy of the questionnaire and a consent form and asked to return both
when they had been completed. For the purposes of evaluating test-retest reliability,
parents were sent a second questionnaire approximately 2 weeks after completing the
first. Formal carers were identified through a national long-stay
residential/assessment centre.
The mean age of children recruited to the study was 11.5 years (SD 4.6, range 2 –
191); 58% were male and 42% female. Of the parent respondents, 39 were mothers, 3
were fathers, and 2 were other relatives; 3 of the questionnaires were jointly
completed by both parents.
The final version consisted of 70 items covering seizure severity, seizure-related
injuries, AED side-effects, behaviour, mood, physical functioning, cognitive
functioning, social functioning, parental concern, communication, overall QOL and
overall health. The items covering behaviour, seizure severity, mood and side-effects
domains are summed to create 4 subscales containing 9, 14, 16 and 19 items
respectively. On each subscale, a higher score indicates poorer functioning. The
remaining domains consist of single or several non-summed items. Subscales from
two other outcome measures [the Aberrant Behavior Checklist (ABC)16 and the Child
Health Questionnaire (CHQ)]17 were also administered for the purposes of evaluating
the construct validity of ELDQOL.
In order to evaluate validity, t-tests and correlations were used to examine the
relationship between ELDQOL sub-scales and:
* the Irritability and Hyperactivity subscales of the ABC;
* the Impact on Parental Time, Emotional Impact on Parent, and Family
Activities sub-scales of the CHQ;
* 4 global items (overall health, overall QOL, perceived severity of condition 1 Age was not recorded for 13 children.
6
D Buck
and disability level).
Severity of disability was assessed using a 4-item scale developed by a consultant
paediatric neurologist (RA) with a specialist interest in epilepsy and learning
disability. It covers mobility, feeding, dressing and speech, each item having 5
response options ranging from the worst to the best possible level of functioning.
It was predicted a priori that there would be:
* a significant relationship between the ABC Irritability and Hyperactivity
scales and scores on each ELDQOL subscale;
* a significant relationship between the 3 CHQ scales and scores on each
ELDQOL subscale;
* a significant relationship between overall health and scores on each ELDQOL
subscale;
* A significant relationship between overall QOL and scores on the Side-Effects,
Seizure Severity, and Mood subscales
* A significant relationship between perceived severity of condition and each
subscale
* A significant relationship between disability level and Seizure Severity.
Internal consistency of the 4 sub-scales was assessed using Cronbach’s alpha
coefficients and item-total correlations. Test-retest reliability was assessed using
intraclass correlation coefficients. Floor and ceiling effects for each subscale were
examined.
7
D Buck
Results
Qualitative phase
Views of parents
Six issues were identified through the cognitive interviews with parents: validity of
domains, multiple seizure types, wording, ease of completion, response categories and
identifying change. It was clear that the questionnaire resonated with parents’
experiences and concerns, the questions being described as ‘very good’ and ‘geared to
parents’ for example. However, it did not, in its existing form, take into account that
some children have no mobility or verbal skills. Some commented on the lack of
lifestyle questions to capture social activities and interests. In terms of multiple
seizure types, some parents found it difficult to identify which seizure type they
should refer to in answering the questions. There was a consensus that use of the
word ‘seizure’ was preferable to ‘attack’. The ‘layman’s language’ was appreciated
because it was easy to understand, and the length of the questionnaire was viewed as
good, ‘quick to do’ and ‘easily to fill in’. Instructions were described as ‘clear’,
‘logical’ and ‘concise’. Overall, parents felt that they had sufficient choice in
response categories, and being asked to concentrate on the past 4 weeks seemed
acceptable.
Views of professionals
Interviews with professionals highlighted several broad methodological issues, in
particular the nature of the proxy respondent, scope and objectives of the
questionnaire, sensitivity to change, and questionnaire design. Participants felt
generally that parents would be able to answer the majority of the questions and
would therefore be appropriate proxies. However, there was some reservation
regarding the suitability of nursing, teaching or respite care staff who may not have
known the child for long – ‘it would be necessary to know the child inside out’ to be
able to complete some of the sections of the questionnaire.. In terms of the scope and
objectives of the questionnaire, its orientation led some to query what was being
sought and by whom it would be used. The addition of socially based QOL questions,
8
D Buck
a ‘health profile’ and a section covering function was suggested. In terms of design,
most professionals felt the length of the questionnaire was about right. Some thought
the use of the word ‘attack’ was regressive in terms of practice and preferred the term
‘seizure’.
Revisions made as a result of the qualitative phase
Taking into account the comments from parents and professionals, a number of
revisions were made. In summary:
* the word ‘seizures’ replaced the word ‘attacks’
* ‘can’t say’ and ‘does not apply’ response options were added
* a question about perceived control was dropped because of ambiguity of
meaning; and one about perceived severity was added
* an item was added to those about seizure-related injuries
* an item covering ‘usual activities’ (school, playing, socialising) was added to
the behaviour subscale
* 2 items (‘sociable’ and ‘depressed’) were dropped from the mood subscale and
replaced by three new ones (‘sad’, ‘withdrawn’, ‘cooperative/helpful’)
* 2 items were dropped from the side effects profile (hair loss, dizziness) and 2
were added (weight loss, loss of appetite); and a single item,
‘restlessness/hyperactivity’ was separated into two
* clearer/more detailed instructions were provided (including a statement
instructing parents/proxies to think about the most severe seizures in the case
of multiple seizure types, and emphasising that questions relate to the last four
weeks)
9
D Buck
Psychometric phase
Reliability
As shown in Table 1, internal consistency of the Behaviour, Seizure Severity, Mood
and Side-Effects scales was high (0.74 – 0.95). Item-total correlations exceeded the
desired 0.4 for most items, although in each scale there were some items which did
not meet this: 3/9 items in the Behaviour Scale (‘problems sleeping‘, ‘being prevented
from taking part in normal activities‘, and ‘appetite a problem‘); 7/14 items in the
Seizure Severity scale (‘aware of surroundings after seizure‘, ‘blank out/lose
consciousness‘, ‘fall to the ground‘, ‘soil self‘, ‘injure mouth/cheek /tongue‘, ‘upset
by injuries‘, ‘appear sleepy/subdued‘); 4/16 items in the Mood scale (‘tearful‘,
‘hyperactive‘, ‘withdrawn‘, ‘cooperative‘); and 3/19 items in the Side-effects scale
(‘skin problems‘, ‘shaky hands‘, ‘weight gain.
The test-retest reliability of each subscale was high (range 0.80 - 0.96, Table 1).
Floor/ceiling effects were acceptable in the Side Effects scale, low in the Seizure
Severity scale, and there were none in the Mood or Behaviour scales (Table 1).
Validity
Table 2 shows that all but one of the 4 ELDQOL subscales (Behaviour) were
moderately to highly correlated with both the Irritability and Hyperactivity scales of
the ABC, and that all 4 were moderately to highly correlated with the Emotional
Impact and Family Activities scales of the CHQ. Two of the ELDQOL subscales
(Seizure Severity and Side-effects) correlated moderately with the Impact on Parental
Time scale of the CHQ.
Table 3 shows a significant relationship between mean scores on the Behaviour and
Mood subscales by overall health. There was a similar though non-significant trend
for the Seizure Severity and Side Effects subscales.
As can be seen in Table 4, those with poorer perceived overall QOL tended to have
10
D Buck
higher mean scores on the ELDQOL scales, although these were not statistically
significant.
Mean scores on each of the 4 subscales were significantly worse for parents who felt
their child’s condition was very severe compared to those who felt it was somewhat or
moderately severe (Table 5).
The Seizure Severity and Behaviour subscales were significantly correlated with
severity of disability (Table 6). No significant correlation was found for the Side-
Effects and Mood subscales.
11
D Buck
Discussion
ELDQOL was derived using currently accepted standards for scale development,
including qualitative and cognitive interview techniques with the target groups. The
ongoing involvement of parents and professionals in determining the content and
format of the measure has enhanced its acceptability and ease of completion. The final
version of ELDQOL has very good evidence of reliability and validity, making it a
promising proxy instrument for assessing the QOL of children/young adults with
epilepsy and learning disability. Although in this validation exercise parents were not
asked to note length of time taken to complete ELDQOL, this was previously
estimated as only around 20 minutes5, a factor which may further contribute to its
acceptability.
In terms of internal consistency, each of the 4 subscales had high Cronbach’s alpha
values. For each scale there were some items which did not meet the desired 0.4 level
for item to total scale correlation.19 However, in each case the effect on Cronbach’s
alpha when these items are removed from the scale is negligible or has no effect and
we therefore propose to retain these items on the grounds of
comprehensiveness/content validity.
The test-retest exercise found that each of the 4 subscales achieved ICCs well above
the required level of 0.70 for group comparisons, and 3 of the 4 performed well
enough for comparing individuals over time (0.90).19 Each scale performed well or
perfectly in terms of floor/ceiling effects.
We found excellent evidence of construct validity as all but one of the 4 ELDQOL
subscales (Behaviour) were moderately to highly correlated with both the Irritability
and Hyperactivity scales of the ABC, and all were moderately to highly correlated
with the Emotional Impact and Family Activities scales of the CHQ. Poorer
perceived health was significantly related to poorer mean scores on two of the
ELDQOL subscales, and a similar trend was seen for the remaining two, providing
further evidence of construct validity. Greater perceived severity of the condition was
significantly related to higher scores on each subscale, and disability level was
significantly correlated with scores on the Seizure Severity and Behaviour scales.
12
D Buck
Although the sample size was relatively small we would suggest that, based on both
psychometric analysis and feedback from parents, ELDQOL will be a valuable
contribution to outcomes measurement in childhood epilepsy and learning disability,
particularly in the context of treatment trials, but also, with the addition of new items
to address parent and professional concerns over coverage and content validity, in a
wider research context. Further studies are needed to assess its responsiveness to
change.
13
D Buck
Table 1 Reliability and floor/ceiling effects of the 4 ELDQOL subscales
ELDQOL Su
bscale
Cronbach’s
alpha
Item-total co
rrelations
Intraclass
correlation
(test-retest)
Floor / ceilin
g effects (%)
Behaviour 0.79 .24 - .70 0.80 0 / 0
Seizure Sever
ity
0.74 .19 - .54 0.96 2 / 0
Mood 0.86 .21 - .78 0.91 0 / 0
Side-effects 0.95 .33 - .88 0.92 10 / 0
14
D Buck
Table 2 Validity: correlation between the 4 ELDQOL sub-scales and ABC
and GHQ sub-scales1
ELDQOL
subscale:
ABC subscales CHQ subscales
Irritability Hyperactivity Impact on
Parental
Time
Emotional
Impact on
Parent
Family
Activities
Behaviour r=0.008
p=0.95
r=0.13
p=0.32
r=0.25
p=0.1
r=-0.42
p<0.01
r=-0.31
p<0.05
Seizure
Severity
r=0.29
p<0.05
r=0.38
p<0.01
r=-0.33
p<0.05
r=-0.47
p<0.01
r=-0.48
p<0.01
Mood r=0.74
p<0.001
r=0.56
p<0.001
r=-0.28
p=0.06
r=-0.48
p<0.01
r=-0.63
p<0.001
Side-effects r=0.38
p<0.02
r=0.39
p=0.01
r=-0.43
p<0.01
r=-0.54
p<0.001
r=-0.47
p<0.01
1 Pearson’s correlation coefficient
15
D Buck
Table 3 Validity: mean scores of the 4 ELDQOL subscales by overall
health
Overall health:
Very good/g
ood
Fair/poor/v po
or
t test p value
Seizure severity 26 31 -1.8 0.08
Side-effects Profi
le
30 37 -1.9 0.06
Behaviour Scale 17 21 -3.6 0.001
Mood 32 35 -2.1 0.038
16
D Buck
Table 4 Validity: mean scores of the 4 ELDQOL subscales by overall QOL
Overall QOL:
Very god/go
od
Fair/poor/v po
or
t test P value
Seizure severity 29 30 -0.2 0.81
Side-effects Profi
le
30 36 -1.4 0.16
Behaviour Scale 17 19 -1.8 0.078
Mood 32 35 -1.8 0.071
17
D Buck
Table 5 Validity: mean scores of the 4 ELDQOL subscales by perceived
severity of condition
How severe:
Very Somewhat/moderate/
Mild
t test p value
Seizure severity 35 24 4.3 <.001
Side-effects
Profile
43 32 2.9 <0.01
Behaviour Scale 21 17 3.1 <.01
Mood 37 31 2.4 <.03
18
D Buck
Table 6 Validity: correlation between ELDQOL subscales and disability
level
Scale Pearson’s r p value
Seizure severity -0.43 .001
Side Effects Profile 0.02 .93
Behaviour scale 0.44 .000
Mood -0.1 .41
19
D Buck
Acknowledgements
We would like to thank all the parents who helped us at the various stages of this
study; and the health professionals who contributed to the revalidation exercise. Our
sincere thanks go to: Drs Tim Betts, Lynda Brook, Richard Newton, Rosemary
Newton, William Whitehouse and Chris Verity; to epilepsy specialist nurses, Lynne
Clack and Gerry Litherland; and to Helen Donovan, for their help with identifying
potential study participants. The study was funded by GSK in its various incarnations,
to whom we also extend our thanks.
20
D Buck
References
1. Cowan J, Baker GA. A review of subjective impact measures for use with
children and adolescents with epilepsy. Qual Life Res 2004 (8):1435-1443.
2. Buck D, Jacoby A, Massey A, Ford G (2000) Evaluation of measures used to
assess quality of life after stroke. Stroke 2000;8:2004-2010.
3. Fallowfield L. The Quality of Life: The Missing Measurement in Health Care.
London, UK : Souvenir Press, 1990.
4. Baker GA, Camfield C, Camfield P et al. Commission on Outcome
Measurement in Epilepsy, 1994-1997: final report. Epilepsia 1998;39(2):213-
231.
5. Espie CA, Kerr M, Paul A et al. Learning disability and epilepsy. 2, A review
of available outcome measures and position statement on development
priorities. Seizure 1997;6(5):337-350.
6. Espie CA, Watkins J, Duncan R. Development and validation of the Glasgow
Epilepsy Outcome Scale (GEOS): a new instrument for measuring concerns
about epilepsy in people with mental retardation. Epilepsia 2001;42(8):1043-
1051.
7. Ronen GM, Streiner DL, Rosenbaum P; Canadian Pediatric Epilepsy Network.
Health-related QoL in children with epilepsy: development and validation of
self-report and parent proxy measures. Epilepsia 2003;44(4):598-612.
8. Camfield C, Breau L, Camfield P. Assessing the impact of pediatric epilepsy
and concomitant behavioral, cognitive, and physical/neurologic disability:
Impact of Childhood Neurologic Disability Scale. Dev Med Child Neurol.
2003 Mar;45(3):152-159.
9. Baker GA, Jacoby A. Quality of life assessment with add-on Lamictal
(Lamotrigine) in children and adults with epilepsy and learning disability.
21
D Buck
Epilepsia 1997;38(Suppl 3):108.
10. Baker GA, Jacoby A, Douglas C. Quality of life outcomes of lamotrigine used
in the treatment of patients with severe epileptic syndromes. Epilepsia
1996;37(suppl 4):117.
11. Jacoby A, Baker G, Bryant-Comstock L, Phillips S, Bamford C. Lamotrigine
add-on therapy is associated with improvements in mood in patients wirh
severe epilepsy. Epilepsia 1996;37(Suppl 5):S202.
12. Jacoby A, Baker GA, Appleton R. Quality of life of children with epilepsy.
Findings from a UK community study. Epilepsia 1996;37(Suppl 4):100.
13. Baker G, Jacoby A, Berney T, Dewey M, Hosking G et al. Development of an
instrument to assess quality of life in children with epilepsy and learning
disability. Epilepsia 1994; 35(Suppl 7): S47.
14. Donaldson JA, Glauser TA and Olberding LS. Lamotrigine adjunctive therapy
in childhood epileptic encephalopathy (the Lennox Gastaut syndrome)
Epilepsia 1997;38:68-73.
15. Buck D, Jacoby A, Massey A, Ford G. Pre-testing items in a stroke-specific
quality of life measure: the value of cognitive interviewing. Journal of
Neurolinguistics 2000; 13: 267-71.
16. Aman MG, Singh NN, Stewart AW, Field CJ. The aberrant behavior checklist:
a behavior rating scale for the assessment of treatment effects. Am J Ment Defi
c 1985;89(5):485-491.
17. Langraf NJ, Abetz L, Ware JE. The child health questionnaire (CHQ): A user's
manual. Boston : The Health Institute, New England Medical Center, 1996.
18. Streiner DL, Norman GR. Health Measurement Scales. A Practical Guide to T
heir Development and Use. Oxford : Oxford University Press, 1995.
22